Prosodic information bootstraps language learning in infancy

韵律信息引导婴儿期的语言学习

• 批准号: RGPIN-2018-03797
• 负责人: Curtin, Suzanne
• 金额: $ 2.4万
• 依托单位: University of Calgary
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=648440
• 关键词: Prosodic; information; bootstraps; language; learning

Language acquisition is one of the hallmark accomplishments of human development. To successfully acquire language, a child must learn about word meaning (semantics) and how words combine to create sentences (syntax). Prior to this, infants are learning about sounds and sound combinations, recognizing some highly frequent word forms, finding words in speech stream, and building representations. Remarkably, infants acquire a great deal of their language within the first 2 years of life. This complex task is supported by the rich input and statistical mechanisms as the language learning system begins to store and organize information.***Our PRIMIR framework (Processing Rich Information from Multidimensional Interactive Representations: Curtin et al., 2011; Werker & Curtin, 2005) includes representational spaces for storing information, learning mechanisms for altering extant representations and building new ones, and dynamic filters (perceptual biases, task demands, and developmental level) that direct information processing. Building on this framework, my NSERC research tests the need to incorporate a prosodic bootstrapping mechanism working in conjunction with general learning mechanisms to facilitate word learning. Specifically, one way for the infant to 'bootstrap' acquisition is by attending to highly salient information in speech. Prosodic information, such as word stress and rhythm, can highlight linguistic structure such as words, their grammatical categories, and sentence boundaries. The overarching goal of my research is to test how prosodic information - specifically word stress - facilitates the detection of linguistic structure in the speech input and bootstraps language learning.***A comprehensive exploration of the processes underlying word learning involves consideration of multiple sources of influence, from basic perception of speech, to the use of complex structures to establish word-referent relationships, such as infants' use of sentences to glean grammatical function to determine word meaning. The proposed experiments reflect these multiple levels of influence on early language learning. The findings from this research program will be of interest to cognitive psychologists, developmental psychologists, and psycholinguists. It will be of particular interest to NSE scientists because understanding speech development is necessary for discovering the neurological and physiological underpinnings of the behaviour and for the advancement speech technologies, such as speech recognition systems.

语言习得是人类发展的标志性成就之一。为了成功地掌握语言，孩子必须学习单词的含义（语义）和单词如何联合收割机组成句子（句法）。在此之前，婴儿正在学习声音和声音组合，识别一些高频率的单词形式，在语音流中找到单词，并建立表征。值得注意的是，婴儿在出生后的头2年内获得了大量的语言。当语言学习系统开始存储和组织信息时，这个复杂的任务得到了丰富的输入和统计机制的支持。我们的PRIMIR框架（Processing Rich Information from Multidimensional Interactive Representations：科廷等人，二〇一一年; Werker &科廷，2005）包括存储信息的表征空间，改变现有表征和构建新表征的学习机制，以及指导信息处理的动态过滤器（感知偏差，任务要求和发展水平）。在这个框架的基础上，我的NSERC研究测试了是否需要将韵律引导机制与一般学习机制结合起来，以促进单词学习。具体来说，婴儿“引导”习得的一种方式是注意语音中高度突出的信息。韵律信息，如单词重音和节奏，可以突出语言结构，如单词，它们的语法类别和句子边界。我研究的首要目标是测试韵律信息-特别是单词重音-如何促进语音输入中语言结构的检测并引导语言学习。对单词学习过程的全面探索涉及到对多种影响来源的考虑，从对语音的基本感知，到使用复杂的结构来建立单词-所指关系，例如婴儿使用句子来收集语法功能以确定单词含义。这些实验反映了这些对早期语言学习的多层次影响。这项研究计划的结果将引起认知心理学家，发展心理学家和心理语言学家的兴趣。这将是特别感兴趣的NSE科学家，因为理解语音发展是必要的，发现神经和生理基础的行为和语音技术的进步，如语音识别系统。